Measuring coordination between women’s self-help groups and local health systems in rural India: a social network analysis

Objectives To assess how the health coordination and emergency referral networks between women’s self-help groups (SHGs) and local health systems have changed over the course of a 2-year learning phase of the Uttar Pradesh Community Mobilization Project, India. Design A pretest, post-test programme evaluation using social network survey to analyse changes in network structure and connectivity between key individuals and groups. Setting The study was conducted in 18 villages located in three districts in Uttar Pradesh, India. Intervention To improve linkages and coordination between SHGs and government health providers by building capacity in leadership, management and community mobilisation skills of the SHG federation. Participants A purposeful sampling that met inclusion criteria. 316 respondents at baseline and 280 respondents at endline, including SHG members, village-level and block-level government health workers, and other key members of the community (traditional birth attendants, drug sellers, unqualified rural medical providers, pradhans or elected village heads, and religious leaders). Main outcome measures Social network analysis measured degree centrality, density and centralisation to assess changes in health services coordination networks at the village and block levels. Results The health services coordination and emergency referral networks increased in density and the number of connections between respondents as measured by average degree centrality have increased, along with more diversity of interaction between groups. The network expanded relationships at the village and block levels, reflecting the rise of bridging social capital. The accredited social health activist, a village health worker, occupied the central position in the network, and her role expanded to sharing information and coordinating services with the SHG members. Conclusions The creation of new partnerships between traditionally under-represented communities and local government can serve as vehicle for building social capital that can lead to a more accountable and accessible community health delivery system.

Microfinance institutions comprised of Self-Help Groups (SHGs) are increasingly recognized as promising avenues for expanding health and social services to vulnerable populations. [1][2] [3] Activating demand-side interventions through SHGs by providing information on key health practices, strengthening coordination with community health services, and rollingout schemes and services is an important strategy in reaching vulnerable groups.
Other research demonstrates how increasing poor women's access to working capital can result in improvements in education and health [4] [2] as SHGs tend to use their savings and credit for the family's well-being. Enhancing a woman's agency sets into motion new abilities to "exercise bargaining power as well as develop a sense of self-worth, a belief in one's ability to secure desired changes, and the right to control one's life". [5] As women in SHGs build social capital, they also act as an instrument for providing meaningful input into local governance to address deficits in government health systems. [6] [7] Building on the strengths of SHGs  and Block Organisations (BOs). Through capacity building and leadership training, women who were poor and from lower castes were encouraged to access financial products, as well as expand their knowledge and ability to advocate for government provided entitlements. The objective of our study was to assess how health services coordination and referral networks between SHGs and local health systems along with other key stakeholders changed over the course of a two-year Learning Phase of the project using social network analysis.

Relevance of coordination networks
Coordination among community institutions is achieved through partnerships to improve responses to public and social issues [8] and are built around norms of reciprocity and trustworthiness. [9] The effectiveness of social networks is dependent on the density of community connections and the vibrancy of associations [10] to expand the relationships between diverse groups and to obtain the full range of knowledge, skills, and resources that the community needs to solve complex problems. Bridging social ties are most effective between people and organisations from typically under-represented community members and people with various kinds of expertise that provide access to schemes and services (such as frontline workers, doctors, etc.). [11]

SHG's role in coordination networks
SHG platforms have reached 57.9% of villages in India resulting in 4.8 million credit-linked groups in 2010, which demonstrates the broad potential for generating an empowered community voice to demand accountability from government functionaries. [1] Many social service models, such as the UPCMP aim to promote coordination with government health and social services to expand the exchange of resources and generate social capital. [12] [13] [14] [15] [16] When community networks such as SHGs expand by linking with government programs and providers, more resources are available and together, these groups can tackle issues that no one group can resolve by itself. [17] Participatory policies and community participatory initiatives have been well studied and critiqued. [18] [19] [20] However, studies focusing on how economically marginalized women engage through SHGs in coordinating with government departments are difficult to find in the published literature. Our study aims to contribute to this research gap by examining the efforts of SHGs in coordinating the delivery of health services with the local government health system, to build and sustain effective networks that enable the flow of resources and services to the poorest communities.

Setting
Uttar Pradesh (UP), at approximately 200 million population (India Census, 2011) [21] is one of the largest states in India constituting 16.5% of the country's total population. UPCMP (GPs) located in ten blocks and eight districts with a goal to scale-up the intervention to 120 blocks over the five year project period. During the Learning Phase of the project, and after reviewing the characteristics of each district, we selected one block from three districts that represented different parts of the state (Hardoi, Mirzapur, and Banda) ( Figure   1). In each of the selected blocks, we then selected 6 out of the 10 GPs, for a total of 18 GPs in our analysis.

Linkages intervention
The program's linkages strategy aimed to improve coordination between SHGs and government health and social services by expanding the leadership, management and community mobilization skills of the SHG federation at the Village Organisation (VO) and

Study design
We used a pre-test, post-test social network survey and added qualitative in-depth interviews in the post-test period. The study objective was to assess how the health coordination and referral networks between women's Self Help Groups (SHGs) and local health systems have changed over the course of a two-year Learning Phase of the Uttar Pradesh Community Mobilization Project (UPCMP). The survey instruments were based on a validated SNA survey design [22] and questions were adapted to capture aspects of village and block level connections that would be relevant for assessing the program. Qualitative interviews were developed in the endline to complement the SNA in understanding the changing roles and relationships in village health activities that occurred during the implementation period.
As SHGs form in a village, two members from each SHG are voted into a Village Organization (VO), whose members represent on average about 150-250 SHG women and subsequently, two members from each VO across many villages are elected to the BO at the block level representing 5,000-7,000 women. At the Village level, VO members' main focus was to establish functional relationships between SHG federation members and the three local government health worker cadres designated as "AAAs": the Accredited Social Health Activist (ASHA), a community health worker who gets paid based on her ability to mobilize pregnant and recently delivered women to seek recommended health services and promote institutional deliveries; the Auxillary Nurse Midwife (ANM), a trained midwife who organizes monthly health clinics in each village and supervises the ASHA; and the Anganwadi Worker (AWW), a local nutrition worker who operates a day care center and distributes supplementary food for eligible children and pregnant and lactating women. At the Block level, SHG members voted into BO leadership roles were tasked with developing relationships with the supervisors of the village level health workers, medical staff working at the primary health care centers and other block level government functionaries and elected office holders.

Study sample
The sociometric data was collected through a purposeful sampling methodology that included SHG members and those participating at the federated VO and BO levels, government and private health workers, RGMVP staff and other key stakeholders who were involved in referrals and provision of health advice and services. Block level respondents were only interviewed about their relationship with respondents from two of the six GPs included in the study to reduce the length of the interviews for block level respondents. The total sample was 596, with 316 respondents in the baseline and 280 in the endline for response rates of 94% and 82%, respectively. Table 1 presents the complete list of respondents along with their acronyms used in SNA visual plot construction, village or block location, and a brief summary of their respective roles. The respondents were grouped into four broad categories corresponding to their affiliation. They included: SHG Structure, RGMVP Staff, Government Health & Nutrition, and "Other" key stakeholders. This approach incorporates elements of both a relational and positional models for examining networks of relations with the network structure. [23]  The middle tier of the system of local self-government in India operating at administrative level of block that links villages with Districts * Technical Support Unit (TSU) project workers: TSUC -TSU Community Specialist; TSUN -TSU Nurse Mentor; TSUB -TSU Block Coordinator

Data collection
The baseline data was collected between November 2013 -January 2014 and endline surveys were adminstered between October -November 2015. Surveys were developed in CSPro 6.0 and adminstered using electronic tablets. During both data collection periods, the survey team spent two weeks in each of the three survey districts. Each team had a supervisor that monitored data quality and a UPCMP project representative that provided logistical support.

Analytic structure and network measures
In consideration of the models to guide network measurement and analysis at the individual and whole network level, our work fits into  categorization including: degree centrality, density, and centralization. [11] These measures level and between key players comprising GP-block relationships. Degree centrality measures the number of connections or ties that each respondent maintains [24] and our analysis is based on a mutual confirmation process. In other words, if one person acknowledges a relationship and the other person does not, that relationship is dropped from the network. Density is often used as a measure of social capital [9] and is defined as a ratio of existing relationships or ties in comparison to the potential number of linkages. [25] Centralization is an expression of how tightly the network structure is organized around its most central point. The general procedure is to calculate the differences between the centrality scores of the most central point and those of all other points to generate a ratio of the actual sum of differences to the maximum possible sum of differences. [25]

Data analysis
Social network analysis methods have been used to study the structural makeup of cooperation that can lead to stronger collaborative relationships. [26] [27] The analysis utilized a combination of two software tools: "R" Version 3 (www.r-project.org) [28] and UCINET Version 6. [29] The plot visualization was developed by using NetDraw. [30] Results

Descriptive
The characteristics of study respondents for the two study periods are presented in Table   2. There were no significant differences between the study respondents for baseline and 13 endline in the sociodemographic characteristics except for the affiliation category. There was a decrease of 21 respondents in the SHG group in the endline as reflected in the reduction from 38% of overall respondents to 35%. The median age of the respondents was approximately 40 years. About a quarter to a third of all respondents were not educated. In both rounds of the survey, more than 50% of SHG-level respondents reported having no formal education, while all government health workers reported some education and approximately one third reported post-graduate studies. There was a reduction of scheduled caste and scheduled tribe and an increase in backward caste representation in the endline (see Table 2 footnote). Four district level respondent roles were added in the endline.  (12) 203 (73) 34 (12) 10 (4)

TOTAL 316 280
The classification of castes is formalized by the Government of India into these categories and we used standard definitions to create these categories. Some of these lower caste designations enable caste groups to receive specific government benefits and subsidies. * Chi Square test: p=0.05 Respondents were also asked whether they had friends or neighbors who were SHG members, and these results are presented in Table 3. The percentage of SHGs in the federated structure that knew other SHG members between the baseline and endline increased significantly from 87% to 100%. The majority of RGMVP staff respondents knew SHG members in their GP, or village. Only about one third of government health workers knew an SHG member, while almost two thirds of respondents from the "Other" category (including the Pradhan, Drug Shop Owner, Traditional Birth Attendant etc.) knew an SHG member in their GP.   Table 4). However, since block level respondents were asked about relationships with only two GPs, we confirmed whole GP-block relationships for these two GP-block dyads in each of the three blocks along with results for baseline and endline ( Table 5).

Network metrics
Density of the overall village health services and referral networks in 18 GPs (villages) were low but increased from baseline to endline in two out of three districts: Banda (7.5% to 10.1%) and Mirzapur (4.7% to 12.6%). Hardoi had a higher density at baseline compared to the other two blocks and had a slight reduction at endline (13.6% to 11.4%). While the average pattens of density changes in blocks during the two year intervention period increased, variability was noted beween individual villages in each of the blocks related to their baseline characteristics, and the degree of change that occurred between baseline and endline. Although Mirzapur's density was the lowest in the GPs measured, they had the greatest growth, as noted by tripling of their scores during the two year UPCMP intervention. Hardoi had a reduction in density in half of GPs but remained within a close range to the other two blocks at endline. The GP-block level health services and referral networks take into account the relationships between two villages and block level respondents within each block ( Table 5).
The baseline densities were lower overall than village level networks but increased after two years of project implementation, although the level of increase was less than in the village level networks. Similar to village level network patterns, Hardoi's GP-block level health services and referral network remained at the highest level in comparison to the other two blocks in both baseline and endline, but decreased slightly in the endline (    At baseline, the Banda GP-Block level health services and referral network was organized into two small worlds, with the SHG structure and the RGMVP SHG support structure in one cluster while the government health system and a few other key village respondents were in another separate cluster (Figure 3a). Within the SHG small world cluster, RGMVP was at the center of a centralized system resembling a "spoke and wheel" structure with members of the SHG structure at the village and block levels connecting to RGMVP but not to each other.
In the second cluster, the ASHA was a leading connector of village level respondents (AWW, the rural medical practitioner (RMP), an unqualified provider and an SHG member) to the broader government health system at the Block level (ANM supervisor, primary health center medical officer (PHM) and the village oriented ANM provider. By the endline, the two separate clusters formed one network (Figure 3b)

Limitations
Our study has limitations related to the constraints of social network analysis, limited prior research on coordination network between SHGs and government and limited duration of testing of the full effects of the lingages intervention. SNA has been mainly used for descriptive purposes and it has been less frequently used for evaluating interventions [31] especially in settings that aim to understand the connections between marginalized groups and the government. Therefore this study cannot be generalized.
Sources of bias may include using purposive sampling to identify the respondnet roles in the villages and blocks. Additionally, there are no established criteria for evaluating networks [32] as SNA represents a relatively new multi-disciplinary methodology with limited empirical studies. There is evidence that density and network centrality constructs may influence partnership functioning that affects coordination, but their magnitudes and mechanisms of effect in public health are still largely unknown. [11] Discussion After two years of UPCMP project implementation, the health services coordination and referrals networks increased in their density but absolute levels remained low. This was a positive direction as higher density is considered an overall indicator of cohesion and interaction within a network and is often associated with faster rates of information diffusion within a community. [9] As information is more quickly shared, processed and appropriated, it can lead to shared decision making. [33] Our study demonstrated an increase in connections not only between individuals but also an expansion of relationships between groups. Poor and traditionally lower-caste women formed SHGs and developed skills to cross boundaries and develop relationships with health providers. A strategy based on increasing diversity in relationships while actively working to reduce redundancy can lead to improved levels of bridging social capital in a network. Such an approach that connects essential groups can lead to better ways of coordinating and collaborating. [9] As proxies for bridging social capital, the average degree centrality increased by creating new connections within the two years of UPCMP implementation. The growth in the number of ties from the GP health system to the SHG platform signaled a visible shift to improved communication to coordinate community health services and referrals.

Instructions to authors
Complete this checklist by entering the page numbers from your manuscript where readers will find each of the items listed below.
Your article may not currently address all the items on the checklist. Please modify your text to include the missing information. If you are certain that an item does not apply, please write "n/a" and provide a short explanation.
Upload your completed checklist as an extra file when you submit to a journal.
In your methods section, say that you used the STROBE cross sectional reporting guidelines, and cite them as: Eligibility criteria #6a Give the eligibility criteria, and the sources and methods of selection of participants. Descriptive data #14a Give characteristics of study participants (eg demographic, clinical, social) and information on exposures and potential confounders. Give information separately for exposed and unexposed groups if applicable.

Abstract
Objectives To assess how the health coordination and emergency referral networks between women's Self Help Groups (SHGs) and local health systems have changed over the course of a two-year Learning Phase of the Uttar Pradesh Community Mobilization Project (UPCMP), India.
Design A pre-test, post-test program evaluation using social network survey to analyze changes in network structure and connectivity between key individuals and groups.

Setting
The study was conducted in 18 villages located in 3 districts in Uttar Pradesh,

India.
Intervention To improve linkages and coordination between SHGs and government health providers by building capacity in leadership, management and community mobilization skills of the SHG federation.

Background and Study objective
Microfinance institutions comprised of Self-Help Groups (SHGs) are increasingly recognized as promising avenues for expanding health and social services to vulnerable populations. [1][2][3] In India, the concept of women's SHGs has evolved over the past three decades. Then and now, the basic SHG structure remains defined as informal groups of 10-20 women from similar socio-economic backgrounds living in close proximity. [4] During the 1980s, the objective was to engage women in collective savings activities and provide access to credit. By the early 1990s, the official SHG and Bank Linkage Programme in India was led by the National Agricultural Bank for Rural Development (NABARD) and focused on loans for livelihood activities. To improve scalability, by early 2000s, the SHG model grew into a key government programme providing financial access to the poor and addressing issues of social justice to improve the welfare of its members. [5] The evidence for supporting women's SHGs continues to grow as research demonstrates that increasing poor women's access to working capital can result in improvements in education and health [2,6] and that SHGs tend to use their savings and credit for the family's well-being. Enhancing a woman's agency sets into motion new abilities to "exercise bargaining power as well as develop a sense of self-worth, a belief in one's ability to secure desired changes, and the right to control one's life". [7] As SHGs build social capital, they can also be instrumental in addressing deficits in government health systems.  Phase of the project using social network analysis.

Relevance of coordination networks
Coordination among community institutions is achieved through partnerships to improve responses to public and social issues [10] and are built around norms of reciprocity and trustworthiness. [11] The effectiveness of social networks is dependent on the density of community connections and the vibrancy of associations [12] to expand the relationships between diverse groups and to obtain the full range of knowledge, skills, and resources that the community needs to solve complex problems.
Bridging social ties are most effective between people and organisations from typically under-represented communities and groups with expertise that can provide access to schemes and services (such as frontline workers, and doctors). [

Linkages intervention
The linkages strategy aimed to improve coordination between SHGs and government At Community/Neighborhood Level, 10-20 women from particularly vulnerable and marginalised households are organised into Self-Help Groups. These SHGs meet regularly for the purpose of addressing common problems through mutual-support. In the case of RGMVP, SHGs are responsible for promoting savings among groups, ensuring credit access from banks, and driving community-based social and behaviour change interventions associated with maternal, newborn and child health. As SHGs form in a village, two members from each SHG are voted into a Village Organisation, whose members represent on average about 150-250 SHG women. Subsequently, two members from each Village Organisation across many villages are elected to the Block Organisation at the block level representing 5,000-7,000 women.
At the Village level, Village Organisation members' main focus was to establish functional relationships between SHG federation members and the three local government health worker cadres designated as "AAAs": the Accredited Social Health Activist (ASHA), a community health worker who gets paid based on her ability to mobilize pregnant and recently delivered women to seek recommended health services and promote institutional deliveries; the Auxillary Nurse Midwife (ANM), a trained midwife who organizes monthly health clinics in each village and supervises the ASHA; and the Anganwadi Worker (AWW), a local nutrition worker who operates a day care center and distributes supplementary food for eligible children and pregnant and lactating women.
At the Block level, Village Organization members are voted into the Block Organisation. Their roles were to develop relationships with the supervisors of the village level health workers, medical staff working at the primary health care centres and other block level government functionaries and elected office holders.

Study design
We developed a pre-test, post-test social network survey. The study objective was to assess how the health coordination and emergency referral networks between women's SHGs and local health systems changed over the course of the two-year Learning Phase of the Uttar Pradesh Community Mobilization Project (UPCMP). The survey instruments were based on a validated survey design structure [24] and questions were developed and then pre-tested in a social and cultural setting similar to the study population to capture aspects of village and block level connections that would be relevant for assessing the program. One of the main survey questions asked every respondent whether they coordinated health services, including emergency referrals with every other respondent type in the survey (Table1).

Study sample
The data was collected through a purposeful sampling methodology that included SHG members and those participating at the federated Village Organisation and Block Organisation levels, government and private health workers, RGMVP staff and other key stakeholders. Block level respondents were only interviewed about their relationship with respondents from two of the six Gram Panchayats to reduce the length of the interviews for block level respondents.
Certain roles are unique in a village or block, such as the ASHA, ANM, Pradhan or village leader and for those roles the inclusion criteria were that the respondent agree to the and attempted to contact them in a random order. The first potential respondent that was successfully contacted was interviewed. The total sample was 596, with 316 respondents in the baseline and 280 in the endline for response rates of 94% and 82%, respectively.

Data collection
The baseline data were collected between November 2013 -January 2014 and endline surveys were adminstered between October -November 2015. Surveys were developed in CSPro 6.0 and adminstered using electronic tablets. During both data collection periods, the survey team spent two weeks in each of the three survey districts. Each team had a supervisor that monitored data quality and a UPCMP project representative that provided logistical support.

Analytic structure and network measures
In consideration of the models to guide network measurement and analysis at the individual and whole network level, our work fits into Mays and Scutchfield's (2010) categorization including: degree centrality, density, and centralization. [13] These  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  In other words, if one person acknowledges a relationship and the other person does not, that relationship is dropped from the network. Overall, 31% of ties remained after the confirmation process. The high loss of unconfirmed ties signals a weak level of connectivity whereas ties confirmed by both parties have a higher probability of producing collaborative relationships. [26] In India, the caste system exerts barriers to relationship formation and contributes to a reduction in reciprocity of ties. Density is often used as a measure of social capital [11] and is defined as a ratio of existing relationships or ties in comparison to the potential number of linkages. [27] Centralization is an expression of how tightly the network structure is organized around its most central point. The general procedure is to calculate the differences between the centrality scores of the most central point and those of all other points to generate a ratio of the actual sum of differences to the maximum possible sum of differences. [27]

Data analysis
Social network analysis methods have been used to study the structural makeup of cooperation that can lead to stronger collaborative relationships. [28][29] The analysis utilized a combination of two software tools: "R" Version 3 (www.r-project.org), [30] and UCINET Version 6. [31] The plot visualization was developed by using NetDraw. [32]

Descriptive
The characteristics of study respondents for the two study periods are presented in Table 2

TOTAL 316 280
The classification of castes is formalized by the Government of India into these categories and we used standard definitions to create these categories. Some of these lower caste designations enable caste groups to receive specific government benefits and subsidies. Respondents were asked to self-identify into caste categories in the survey. * Chi Square test: p<0.001 Respondents were also asked whether they had friends or neighbors who were SHG members, and these results are presented in Table 3.

Social Network Analysis
A fixed set of Gram Panchayat and block level providers and other key community members involved in health care delivery were included in the analysis. Network measures for density, centralization and average degree centrality are presented individually for all 18 villages (6 per district) as well as the average scores across each district for the baseline and endline (Table 4). However, since block level respondents were asked about relationships with only two Gram Panchayats, we confirmed whole Gram Panchayat-block relationships for these two Gram Panchayat-block dyads in each of the three blocks along with results for baseline and endline ( Table 5). Panchayats (villages) were low but increased from baseline to endline in two out of three districts: Banda (7.5% to 10.1%) and Mirzapur (4.7% to 12.6%). Hardoi had a higher density at baseline compared to the other two blocks and had a slight reduction at endline (13.6% to 11.4%). While the average pattens of density changes in blocks during the two year intervention period increased, variability was noted beween individual villages in each of the blocks related to their baseline characteristics, and the degree of change that occurred between baseline and endline. Although Mirzapur's density was the lowest in the Gram Panchayats measured, they had the greatest growth, as noted by tripling of their scores during the two year UPCMP intervention.

Network metrics
Hardoi had a reduction in density to half of Gram Panchayats but remained within a close range to the other two blocks at endline. The Gram Panchayat-block level health services and referral networks take into account the relationships between two villages and block level respondents within each block ( Table 5). The baseline densities were lower overall than village level networks but

Health services and emergency referral networks
The   At baseline, the Banda Gram Panchayat-Block level health services and referral network was organized into two small worlds, with the SHG structure and the RGMVP SHG support structure in one cluster while the government health system and a few other key village respondents were in another separate cluster (Figure 3a). Within the SHG small world cluster, RGMVP was at the center of a centralized system resembling a "spoke and wheel" with members of the SHG structure at the village and block levels connecting to RGMVP but not to each other. In the second cluster, the ASHA was a leading connector of village level respondents (AWW, the rural medical practitioner (RMP), an unqualified provider and an SHG member) to the broader government health system at the Block level (ANM supervisor, primary health center medical officer (PHM) and the village oriented ANM provider). By the endline, the two separate clusters formed one network (Figure 3b). The ASHA became the main connector between the government health providers and the SHG structure. The SHG Village Organisation members also increased coordination within their own network and RGMVP. RGMVP scaled back from their centralized role at baseline and moved to the periphery of the network while SHG enagagment increased.
In Hardoi, the structure remained similar since they started with one whole network, however an additional project, the Technical Support Unit (TSU), a supply side intervention working with the health system, was just starting during the endline collection period (Figures 4a -4b). The

Limitations
Our study has limitations related to the constraints of social network analysis, minimal prior research on coordination network between SHGs and government and limited duration of testing of the full effects of the lingages intervention. SNA has been mainly used for descriptive purposes and it has been less frequently used for evaluating interventions [33] especially in settings that aim to understand the connections between marginalized groups and the government. Therefore this study cannot be generalized. Sources of bias may include using purposive sampling to identify the respondnet roles in the villages and blocks. Additionally, there are no established criteria for evaluating networks [34] as SNA represents a relatively new multidisciplinary methodology with limited empirical studies. There is evidence that density and network centrality constructs may influence partnership functioning that affects coordination, but their magnitudes and mechanisms of effect in public health are still largely unknown. [13] Discussion After two years of UPCMP project implementation, the health services coordination and emergency referral networks increased in their density but absolute levels remained low. This was a positive direction as higher density is considered an overall indicator of cohesion and interaction within a network and is often associated with faster rates of information diffusion within a community. [11] As information is more quickly shared, processed and appropriated, it can lead to shared decision making. [35] Our study demonstrated an increase in connections not only between individuals but also an expansion of relationships between groups. Poor and traditionally lower-caste SHGs developed skills to cross boundaries and forge relationships with health providers.
Traditionally bound by societal and cultural expectations, it has been difficult to overcome social and structural barriers, which limit interactions among heterogenous groups. A strategy based on increasing diversity in relationships while actively working to reduce redundancy can lead to improved levels of bridging social capital in a network. Such an approach that connects essential groups can lead to better ways of coordinating and collaborating. [11] Therefore, SHGs with increasing skills and confidence in engaging with health services providers are creating in-roads to better health services for themselves and their communities.
As proxies for bridging social capital, the average degree centrality increased by creating new connections within the two years of UPCMP implementation. However, simply increasing the number of ties may not necessarily result in strong relationships, as noted by the weak ties theory. [36] Although it is common to surround ourselves with strong ties that include people similar to us in beliefs, values, and access to resources, it is through weak ties that we begin to diversify our networks and create avenues for accessing more varied resources. [11] It is probable that building SHG leadership and coordination skills, which was a major UPCMP intervention, facilitated a pathway for SHGs to create weak ties with the health system. These patterns reflect opportunities for diverse stakeholders to engage in the planning and production of change. [37] Our analysis showed that the ASHA occupied the central position in the network and her role expanded to coordinating services with the Village Organization Swasthya Sakhi and others in the SHG federation. The ASHA is a community health volunteer and her role evolved from an "activist" and advocate of her village to being part of the government health system and being accountable to the ANM, or nurse midwife as well as the  [11,43] If successful, these partnerships can serve as vehicles for transforming public health from a diverse collection of activities and organisations into an organized and accountable delivery system. [13] The challenge will be in maintaining these networks, so that they remain dynamic and offer new benefits to partners. [24] Future studies that include longitudinal data can provide deeper understanding of the mechanisms by which intersectoral partnerships and community mobilization lead to effective coordination networks to tackle health problems and their socioeconomic determinants.

Instructions to authors
Complete this checklist by entering the page numbers from your manuscript where readers will find each of the items listed below.
Your article may not currently address all the items on the checklist. Please modify your text to include the missing information. If you are certain that an item does not apply, please write "n/a" and provide a short explanation.
Upload your completed checklist as an extra file when you submit to a journal.
In your methods section, say that you used the STROBE cross sectional reporting guidelines, and cite them as: Eligibility criteria #6a Give the eligibility criteria, and the sources and methods of selection of participants.

Abstract
Objectives To assess how the health coordination and emergency referral networks between women's Self Help Groups (SHGs) and local health systems have changed over the course of a two-year Learning Phase of the Uttar Pradesh Community Mobilization Project (UPCMP), India.
Design A pre-test, post-test program evaluation using social network survey to analyze changes in network structure and connectivity between key individuals and groups.

Setting
The study was conducted in 18 villages located in 3 districts in Uttar Pradesh,

India.
Intervention To improve linkages and coordination between SHGs and government health providers by building capacity in leadership, management and community mobilization skills of the SHG federation.
Participants A purposeful sampling that met inclusion criteria. 316 respondents at baseline and 280 respondents at endline including: SHG members, village and block level government health workers, and other key members of the community (traditional birth attendants, drug sellers, unqualified rural medical providers, pradhans or elected village heads, and religious leaders).   Phase of the project using social network analysis.

Relevance of coordination networks
Coordination among community institutions is achieved through partnerships that improve responses to public and social issues [10] and are built around norms of reciprocity and trustworthiness. [11] The effectiveness of social networks is dependent on the density of community connections and the vibrancy of associations [12] to expand the relationships between diverse groups and to obtain the full range of knowledge, skills, and resources that the community needs to solve complex problems.
Bridging social ties are most effective between people and organisations from typically under-represented communities and groups with expertise that can provide access to schemes and services (such as frontline workers, and doctors). [

Linkages intervention
The linkages strategy aimed to improve coordination between SHGs and government At Community/Neighborhood Level, 10-20 women from particularly vulnerable and marginalised households are organised into Self-Help Groups. These SHGs meet regularly for the purpose of addressing common problems through mutual-support. In the case of RGMVP, SHGs are responsible for promoting savings among groups, ensuring credit access from banks, and driving community-based social and behaviour change interventions associated with maternal, newborn and child health. As SHGs form in a village, two members from each SHG are voted into a Village Organisation, whose members represent on average about 150-250 SHG women. Subsequently, two members from each Village Organisation across many villages are elected to the Block Organisation at the block level representing 5,000-7,000 women.
At the Village level, Village Organisation members' main focus was to establish functional relationships between SHG federation members and the three local government health worker cadres designated as "AAAs": the Accredited Social Health Activist (ASHA), a community health worker who gets paid based on her ability to mobilize pregnant and recently delivered women to seek recommended health services and promote institutional deliveries; the Auxillary Nurse Midwife (ANM), a trained midwife who organizes monthly health clinics in each village and supervises the ASHA; and the Anganwadi Worker (AWW), a local nutrition worker who operates a day care center and distributes supplementary food for eligible children and pregnant and lactating women.
At the Block level, Village Organization members are voted into the Block Organisation. Their roles were to develop relationships with the supervisors of the village level health workers, medical staff working at the primary health care centres and other block level government functionaries and elected office holders.

Study design
We developed a pre-test, post-test social network survey. The study objective was to assess how the health coordination and emergency referral networks between women's SHGs and local health systems changed over the course of the two-year Learning Phase of the UPCMP. The survey instruments were based on a validated survey design structure [24] and questions were developed and then pre-tested in a social and cultural setting similar to the study population to capture aspects of village and block level connections that would be relevant for assessing the program. One of the main survey questions asked every respondent whether they coordinated health services, including emergency referrals, with every other respondent type in the survey (Table1).

Study sample
The data were collected through a purposeful sampling methodology that included SHG and attempted to contact them in a random order. The first potential respondent that was successfully contacted was interviewed. The total sample was 596, with 316 respondents in the baseline and 280 in the endline for response rates of 94% and 82%, respectively.

Data collection
The baseline data were collected between November 2013 -January 2014 and endline surveys were adminstered between October -November 2015. Surveys were developed in CSPro 6.0 and adminstered using electronic tablets. During both data collection periods, the survey team spent two weeks in each of the three survey districts. Each team had a supervisor that monitored data quality and a UPCMP project representative that provided logistical support.

Analytic structure and network measures
In consideration of the models to guide network measurement and analysis at the individual and whole network level, our work fits into  categorization including: degree centrality, density, and centralization. [13] These measures are used to assess changes in health services coordination networks within the Gram Panchayat level and between key players comprising Gram Panchayat -Block  [26] and our analysis is based on a mutual confirmation process.
In other words, if one person acknowledges a relationship and the other person does not, that relationship is dropped from the network. Overall, 31% of ties remained after the confirmation process. The high loss of unconfirmed ties signals a weak level of connectivity whereas ties confirmed by both parties have a higher probability of producing collaborative relationships. [26] In India, the caste system exerts barriers to relationship formation and contributes to a reduction in reciprocity of ties. Density is often used as a measure of social capital [11] and is defined as a ratio of existing relationships or ties in comparison to the potential number of linkages. [27] Centralization is an expression of how tightly the network structure is organized around its most central point. The general procedure is to calculate the differences between the centrality scores of the most central point and those of all other points to generate a ratio of the actual sum of differences to the maximum possible sum of differences. [27]

Data analysis
Social network analysis methods have been used to study the structural makeup of cooperation that can lead to stronger collaborative relationships. [28][29] The analysis utilized a combination of two software tools: "R" Version 3 (www.r-project.org), [30] and UCINET Version 6. [31] The plot visualization was developed by using NetDraw. [32]

Descriptive
The characteristics of study respondents for the two study periods are presented in Table 2

TOTAL 316 280
The classification of castes is formalized by the Government of India into these categories and we used standard definitions to create these categories. Some of these lower caste designations enable caste groups to receive specific government benefits and subsidies. Respondents were asked to self-identify into caste categories in the survey. * Chi Square test: p<0.001 Respondents were also asked whether they had friends or neighbors who were SHG members, and these results are presented in Table 3.

Social Network Analysis
A fixed set of Gram Panchayat and block level providers and other key community members involved in health care delivery were included in the analysis. Network measures for density, centralization and average degree centrality are presented individually for all 18 villages (6 per district) as well as the average scores across each district for the baseline and endline (Table 4). However, since block level respondents were asked about relationships with only two Gram Panchayats, we confirmed whole Gram Panchayat-block relationships for these two Gram Panchayat-block dyads in each of the three blocks along with results for baseline and endline ( Table 5).

Density of the overall village health services and referral networks in 18 Gram
Panchayats (villages) were low but increased from baseline to endline in two out of three districts: Banda (7.5% to 10.1%) and Mirzapur (4.7% to 12.6%). Hardoi had a higher density at baseline compared to the other two blocks and had a slight reduction at endline (13.6% to 11.4%). While the average patterns of density change in blocks during the two year intervention period increased, variability was noted beween individual villages in each of the blocks related to their baseline characteristics, and the degree of change that occurred between baseline and endline. Although Mirzapur's density was the lowest in the Gram Panchayats measured, they had the greatest growth, as noted by tripling of their scores during the two year UPCMP intervention.
Hardoi had a reduction in density to half of Gram Panchayats but remained within a close range to the other two blocks at endline. The Gram Panchayat-block level health services and referral networks take into account the relationships between two villages and block level respondents within each block ( Table 5). The baseline densities were lower overall than village level networks but

Health services and emergency referral networks
The   At baseline, the Banda Gram Panchayat-Block level health services and referral network was organized into two small worlds, with the SHG structure and the RGMVP SHG support structure in one cluster while the government health system and a few other key village respondents were in another separate cluster (Figure 3a). Within the SHG small world cluster, RGMVP was at the center of a centralized system resembling a "spoke and wheel" with members of the SHG structure at the village and block levels connecting to RGMVP but not to each other.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   22 In the second cluster, the ASHA was a leading connector of village level respondents (AWW, the rural medical practitioner (RMP), an unqualified provider and an SHG member) to the broader government health system at the Block level (ANM supervisor, primary health center medical officer (PHM) and the village oriented ANM provider). By the endline, the two separate clusters formed one network (Figure 3b). The ASHA became the main connector between the government health providers and the SHG structure. The SHG Village Organisation members also increased coordination within their own network and RGMVP. RGMVP scaled back from their centralized role at baseline and moved to the periphery of the network while SHG enagagment increased.

Discussion
After two years of UPCMP project implementation, the health services coordination and emergency referral networks increased in their density but absolute levels remained low. This was a positive direction as higher density is considered an overall indicator of cohesion and interaction within a network and is often associated with faster rates of information diffusion within a community. [11] As information is more quickly shared, processed and appropriated, it can lead to shared decision making. [33] However, our study results have limitations related to the constraints of social network analysis, minimal prior research on coordination networks between SHGs and government, and limited duration of testing the full effects of the linkages intervention.
SNA has been mainly used for descriptive purposes and has been less frequently used for evaluating interventions [34] and therefore needs further testing to determine broader generalizability. Sources of bias may include using purposive sampling to identify the respondent roles in the villages and blocks. Additionally, there are no established criteria for evaluating networks [35] as SNA represents a relatively new multi-disciplinary methodology with limited empirical studies. While there is evidence in the literature that the constructs of density and network centrality may influence partnership functioning and coordination, their magnitudes and mechanisms of effect in public health are still largely unknown. [13] Our study provides more empirical evidence to further explore the potential of SNA to measure coordination networks.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   24 Our study demonstrated an increase in connections not only between individuals but also an expansion of relationships between groups. Poor and traditionally lower-caste SHGs developed skills to cross boundaries and forge relationships with health providers.
Traditionally bound by societal and cultural expectations, it has been difficult to overcome social and structural barriers, which limit interactions among heterogenous groups. A strategy based on increasing diversity in relationships while actively working to reduce redundancy can lead to improved levels of bridging social capital in a network. Such an approach that connects essential groups can lead to better ways of coordinating and collaborating. [11] Therefore, SHGs with increasing skills and confidence in engaging with health services providers are creating in-roads to better health services for themselves and their communities.
As proxies for bridging social capital, the average degree centrality increased by creating new connections within the two years of UPCMP implementation. However, simply increasing the number of ties may not necessarily result in strong relationships, as noted by the weak ties theory. [36] Although it is common to surround ourselves with strong ties that include people similar to us in beliefs, values, and access to resources, it is through weak ties that we begin to diversify our networks and create avenues for accessing more varied resources. [11] It is probable that building SHG leadership and coordination skills, which was a major UPCMP intervention, facilitated a pathway for  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   25 SHGs to create weak ties with the health system. These patterns reflect opportunities for diverse stakeholders to engage in the planning and production of change. [37] Our analysis showed that the ASHA occupied the central position in the network and her role expanded to coordinating services with the Village Organization Swasthya Sakhi and others in the SHG federation. The ASHA is a community health volunteer and her role evolved from an "activist" and advocate of her village to being part of the government health system and being accountable to the ANM or nurse midwife as well as the Pradhan (leader) in the Village Health, Nutrition and Sanitation Committee. As an important and recognized member of the community, the ASHA's modeling of crosscaste and class relationships can influence the governance dynamics as building collaborative networks are considered a more democratic means of developing public policy. . [38][39][40][41]

Conclusions
The SHG platform and its federated structure was developed by RGMVP to scale-up and galvanize SHG development to reduce poverty, support women's empowerment, and break caste based hierarchies to encourage comprehensive rural development. The endline social network analysis reveals an expansion of pathways to coordinate health services and emergency referrals for poor, illiterate, and low caste women through the added voices of SHG members. The ability of women with multi-generational social and economic deprivations to broaden their exposure to social networks through SHG  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   26 membership is a major step towards gaining self-confidence in participatory community development.
Based on the STROBE cross sectional guidelines.

Instructions to authors
Complete this checklist by entering the page numbers from your manuscript where readers will find each of the items listed below.
Your article may not currently address all the items on the checklist. Please modify your text to include the missing information. If you are certain that an item does not apply, please write "n/a" and provide a short explanation.
Upload your completed checklist as an extra file when you submit to a journal.

Abstract
Objectives To assess how the health coordination and emergency referral networks between women's Self Help Groups (SHGs) and local health systems have changed over the course of a two-year Learning Phase of the Uttar Pradesh Community Mobilization Project (UPCMP), India.
Design A pre-test, post-test program evaluation using social network survey to analyze changes in network structure and connectivity between key individuals and groups.

Setting
The study was conducted in 18 villages located in 3 districts in Uttar Pradesh,

India.
Intervention To improve linkages and coordination between SHGs and government health providers by building capacity in leadership, management and community mobilization skills of the SHG federation.

Relevance of coordination networks
Coordination among community institutions is achieved through partnerships that improve responses to public and social issues [10] and are built around norms of reciprocity and trustworthiness. [11] The effectiveness of social networks is dependent on the density of community connections and the vibrancy of associations [12] to expand the relationships between diverse groups and to obtain the full range of knowledge, skills, and resources that the community needs to solve complex problems.
Bridging social ties are most effective between people and organisations from typically under-represented communities and groups with expertise that can provide access to schemes and services (such as frontline workers, and doctors). [13]

Linkages intervention
The linkages strategy aimed to improve coordination between SHGs and government At Community/Neighborhood Level, 10-20 women from particularly vulnerable and marginalised households are organised into Self-Help Groups. These SHGs meet regularly for the purpose of addressing common problems through mutual-support. In the case of RGMVP, SHGs are responsible for promoting savings among groups, ensuring credit access from banks, and driving community-based social and behaviour change interventions associated with maternal, newborn and child health. As SHGs form in a village, two members from each SHG are voted into a Village Organisation, whose members represent on average about 150-250 SHG women. Subsequently, two members from each Village Organisation across many villages are elected to the Block Organisation at the block level representing 5,000-7,000 women.
At the Village level, Village Organisation members' main focus was to establish functional relationships between SHG federation members and the three local government health worker cadres designated as "AAAs": the Accredited Social Health Activist (ASHA), a community health worker who gets paid based on her ability to mobilize pregnant and recently delivered women to seek recommended health services and promote institutional deliveries; the Auxillary Nurse Midwife (ANM), a trained midwife who organizes monthly health clinics in each village and supervises the ASHA; and the Anganwadi Worker (AWW), a local nutrition worker who operates a day care center and distributes supplementary food for eligible children and pregnant and lactating women.
At the Block level, Village Organization members are voted into the Block Organisation. Their roles were to develop relationships with the supervisors of the village level health workers, medical staff working at the primary health care centres and other block level government functionaries and elected office holders.

Study design
We developed a pre-test, post-test social network survey. The study objective was to assess how the health coordination and emergency referral networks between women's SHGs and local health systems changed over the course of the two-year Learning Phase of the UPCMP. The survey instruments were based on a validated survey design structure [24] and questions were developed and then pre-tested in a social and cultural setting similar to the study population to capture aspects of village and block level connections that would be relevant for assessing the program. One of the main survey questions asked every respondent whether they coordinated health services, including emergency referrals, with every other respondent type in the survey (Table1).

Study sample
The data were collected through a purposeful sampling methodology that included SHG members and those participating at the federated Village Organisation and Block Organisation levels, government and private health workers, RGMVP staff and other key stakeholders. Block level respondents were only interviewed about their relationship with respondents from two of the six Gram Panchayats to reduce the length of the interviews for block level respondents.

Data collection
The baseline data were collected between November 2013 -January 2014 and endline surveys were adminstered between October -November 2015. Surveys were developed in CSPro 6.0 and adminstered using electronic tablets. During both data collection periods, the survey team spent two weeks in each of the three survey districts. Each team had a supervisor that monitored data quality and a UPCMP project representative that provided logistical support.

Analytic structure and network measures
In consideration of the models to guide network measurement and analysis at the individual and whole network level, our work fits into  categorization including: degree centrality, density, and centralization. [13] These  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   13 measures are used to assess changes in health services coordination networks within the Gram Panchayat level and between key players comprising Gram Panchayat -Block relationships. Degree centrality measures the number of connections or ties that each respondent maintains [26] and our analysis is based on a mutual confirmation process.
In other words, if one person acknowledges a relationship and the other person does not, that relationship is dropped from the network. Overall, 31% of ties remained after the confirmation process. The high loss of unconfirmed ties signals a weak level of connectivity whereas ties confirmed by both parties have a higher probability of producing collaborative relationships. [26] In India, the caste system exerts barriers to relationship formation and contributes to a reduction in reciprocity of ties. Density is often used as a measure of social capital [11] and is defined as a ratio of existing relationships or ties in comparison to the potential number of linkages. [27] Centralization is an expression of how tightly the network structure is organized around its most central point. The general procedure is to calculate the differences between the centrality scores of the most central point and those of all other points to generate a ratio of the actual sum of differences to the maximum possible sum of differences. [27]

Patient and public involvement
No patients or members of the public were involved in the development of research questions, the design of the study, or the development of outcome measures. Also, no patients were asked to advise on interpretation or writing up of results.

Social Network Analysis
A fixed set of Gram Panchayat and block level providers and other key community members involved in health care delivery were included in the analysis. Network measures for density, centralization and average degree centrality are presented individually for all 18 villages (6 per district) as well as the average scores across each district for the baseline and endline (Table 4). However, since block level respondents were asked about relationships with only two Gram Panchayats, we confirmed whole Gram Panchayat-block relationships for these two Gram Panchayat-block dyads in each of the three blocks along with results for baseline and endline ( Table 5).

Network metrics
Density of the overall village health services and referral networks in 18 Gram Panchayats (villages) were low but increased from baseline to endline in two out of three districts: Banda (7.5% to 10.1%) and Mirzapur (4.7% to 12.6%). Hardoi had a  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y 17 higher density at baseline compared to the other two blocks and had a slight reduction at endline (13.6% to 11.4%). While the average patterns of density change in blocks during the two year intervention period increased, variability was noted beween individual villages in each of the blocks related to their baseline characteristics, and the degree of change that occurred between baseline and endline. Although Mirzapur's density was the lowest in the Gram Panchayats measured, they had the greatest growth, as noted by tripling of their scores during the two year UPCMP intervention.
In the second cluster, the ASHA was a leading connector of village level respondents (AWW, the rural medical practitioner (RMP), an unqualified provider and an SHG  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y 22 member) to the broader government health system at the Block level (ANM supervisor, primary health center medical officer (PHM) and the village oriented ANM provider). By the endline, the two separate clusters formed one network (Figure 3b). The ASHA became the main connector between the government health providers and the SHG structure. The SHG Village Organisation members also increased coordination within their own network and RGMVP. RGMVP scaled back from their centralized role at baseline and moved to the periphery of the network while SHG enagagment increased.
In Hardoi, the structure remained similar since they started with one whole network, however an additional project, the Technical Support Unit (TSU), a supply side intervention working with the health system, was just starting during the endline collection period (Figures 4a -4b).
The ANM held the central position in the endline heath services and emergency referral coordination network and acted as an intermediary between the government health functionaries at the block level and the SHG federated structure. The ASHA also stood out as more direct connectivity was established with different members of the SHGs. Within the SHG platform itself, there were more channels of direct communication between different members and women in designated leadership positions. After two years of UPCMP project implementation, the health services coordination and emergency referral networks increased in their density but absolute levels remained low. This was a positive direction as higher density is considered an overall indicator of cohesion and interaction within a network and is often associated with faster rates of information diffusion within a community. [11] As information is more quickly shared, processed and appropriated, it can lead to shared decision making. [33] However, our study results have limitations related to the constraints of social network analysis, minimal prior research on coordination networks between SHGs and government, and limited duration of testing the full effects of the linkages intervention.

Discussion
SNA has been mainly used for descriptive purposes and has been less frequently used for evaluating interventions [34] and therefore needs further testing to determine broader generalizability. Sources of bias may include using purposive sampling to identify the respondent roles in the villages and blocks. Additionally, there are no established criteria for evaluating networks [35] as SNA represents a relatively new multi-disciplinary methodology with limited empirical studies. While there is evidence in the literature that the constructs of density and network centrality may influence partnership functioning and coordination, their magnitudes and mechanisms of effect in public health are still largely unknown. [13] Our study provides more empirical evidence to further explore the potential of SNA to measure coordination networks. Our study demonstrated an increase in connections not only between individuals but also an expansion of relationships between groups. Poor and traditionally lower-caste SHGs developed skills to cross boundaries and forge relationships with health providers.
Traditionally bound by societal and cultural expectations, it has been difficult to overcome social and structural barriers, which limit interactions among heterogenous groups. A strategy based on increasing diversity in relationships while actively working to reduce redundancy can lead to improved levels of bridging social capital in a network. Such an approach that connects essential groups can lead to better ways of coordinating and collaborating. [11] Therefore, SHGs with increasing skills and confidence in engaging with health services providers are creating in-roads to better health services for themselves and their communities.
As proxies for bridging social capital, the average degree centrality increased by creating new connections within the two years of UPCMP implementation. However, simply increasing the number of ties may not necessarily result in strong relationships, as noted by the weak ties theory. [36] Although it is common to surround ourselves with strong ties that include people similar to us in beliefs, values, and access to resources, it is through weak ties that we begin to diversify our networks and create avenues for accessing more varied resources. [11] It is probable that building SHG leadership and coordination skills, which was a major UPCMP intervention, facilitated a pathway for SHGs to create weak ties with the health system. These patterns reflect opportunities for diverse stakeholders to engage in the planning and production of change. [37] 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59

Conclusions
The SHG platform and its federated structure was developed by RGMVP to scale-up and galvanize SHG development to reduce poverty, support women's empowerment, and break caste based hierarchies to encourage comprehensive rural development. The endline social network analysis reveals an expansion of pathways to coordinate health services and emergency referrals for poor, illiterate, and low caste women through the added voices of SHG members. The ability of women with multi-generational social and economic deprivations to broaden their exposure to social networks through SHG membership is a major step towards gaining self-confidence in participatory community development. The increasing success of SHG federatations to forge linkages with the health system led to greater coordination for health services delivery while stimulating a more centralized structure of core village health workers. Collaborative processes that include individual empowerment, bridging social ties, and synergy can strengthen the capacity to solve problems. [42] These ideas feed into the concept of state-society syngery where a mobilized civil society and an active government can work together to build social capital and enhance each other's development efforts. [11,43] If successful, these partnerships can serve as vehicles for transforming public health from a diverse collection of activities and organisations into an organized and accountable delivery system. [13] The challenge will be in maintaining these networks, so that they remain dynamic and offer new benefits to partners. [24] Future studies that include longitudinal data can provide deeper understanding of the mechanisms by which intersectoral partnerships and community mobilization lead to effective coordination networks to tackle health problems and their socioeconomic determinants.

Instructions to authors
Complete this checklist by entering the page numbers from your manuscript where readers will find each of the items listed below.
Your article may not currently address all the items on the checklist. Please modify your text to include the missing information. If you are certain that an item does not apply, please write "n/a" and provide a short explanation.
Upload your completed checklist as an extra file when you submit to a journal.